The Case for Audacity

Florian Neuhaus's assist for Gladbach was stunning. Why don't more players try it?

You have, by now, seen Florian Neuhaus’s pass. I don’t need to describe it to you. It’s the one that made the hairs on your arm snap to attention and salute in the presence of royalty. The one that made you gasp like a lady-in-waiting whose corset is set to “coquettish strangulation.” It is almost certainly the only pass you’ll see this season for which it is not only permissible but legally required that every broadcast angle be spliced together and set to Chopin’s Nocturne in E-flat major, Op. 9, No. 2.

The thing about the pass that makes you feel all fluttery inside is also, in strictly practical terms, the thing that makes it work. The sheer audacity of it, goodness gracious. If one-timing a 60-yard bullet through six defenders were a normal thing to even think of doing, Arturo Vidal might not have been twirling like Julie Andrews in an alpine meadow when it happened. Marcelo Brozović could maybe have intercepted the pass if he hadn’t been checking on two near options, the ones a passer in sound mental health would’ve aimed for. I’m pretty sure poor Danilo D’Ambrosio actually flinched away from the ball. Who could blame him? Moses hid his face from the burning bush, for he was afraid to look upon God.

If we were talking about a shot, this is the part where analytics nerds would step in to politely call you an idiot for experiencing joy at this sport. A highlight-reel goal from distance is thrilling for the same reason it’s dumb. Beating the odds isn’t overthrowing the tyranny of probability; it’s just getting lucky on a bad bet. Beyond a certain distance your team is more likely to score if you do something other than shoot, so getting excited about a long shot that goes in feels like rewarding bad behavior. Shouldn’t we be similarly ambivalent about low-percentage passes?

Like most things in soccer, it depends. The conventional wisdom is that it’s okay to take risks in the final third, since somebody’s eventually going to have to try something to beat the last line of defense, but farther back it’s better to play it safe. Stay calm on the ball. Work the possession forward, don’t force it. There’s a reason center mids are the only position routinely praised for high pass completion percentages.

But what if we’ve got it backwards? Possessions that have already made it to the final third are rare and valuable; show a data analyst the pre-action expected possession value of a ball in the box and she’ll look at you like Adam Sandler pulling a black opal out of a fish. But possessions in the middle third are a dime a dozen: the opportunity cost of trying shit is low, since you probably weren’t going to score anyway, and so is the risk to your defense if you turn it over, since neither are they. If you’ve got any chance at beating the defense from where Neuhaus was, at the bottom of the center circle, it doesn’t have to be much of one for the math to work out in your favor.

Let’s put some numbers on this, just for giggles. Applying event data possession value models to individual plays is bad for the same reason that single-shot xG is bad—that’s not what the model is made for and it doesn’t have enough information, so the precise value won’t mean much—but it might at least help give us some perspective on the problem. In the animation above, each circled number comes from Pieter Robberecht’s online calculator for VAEP, a popular possession value model. According to the calculator, Neuhaus’s successful pass improved his team’s situation by 3.9% of a goal (probably pretty conservative, but never mind that). If the pass had been intercepted by the first line of defense, the last line, or had gone out for a goalkick, its value would have been -0.7%, -0.0%, or -0.2%. Since we don’t know how likely those outcomes are, let’s average them out and say an incomplete pass is worth -0.3%. Work a little middle school algebra on those numbers (+0.039 for a complete pass, -0.003 for incomplete) and you’ll get a break even point—the pass completion probability at which Neuhaus might as well go for the linebreaking assist—of just 7%.

Could Neuhaus have completed that pass 7 times out of 100? I have no idea! I’m not sure I’ve seen 100 pass attempts quite like that in my life, which is exactly what makes it so interesting to me. If a break even success rate of 7% for that pass is even close to the right ballpark, you’d expect to see more players try their luck.

In a talk on possession value, Statsbomb’s Thom Lawrence famously called midfield the “Valley of Meh.” A team in possession isn’t much more likely to score than to concede from the middle third, and most models based on goal probability—most eyeballs, too, for that matter—have trouble detecting the marginal difference between a good or bad action there. That’s a challenge for analytics, but for players the idea that nothing in midfield really matters all that much (because you’re still a number of low-probability contingencies away from goal either way, so whatever) could be liberating. Why not swing for the fences a little more often?

One reason might be psychological. David Romer, the economist who first called attention to the fact that NFL coaches don’t go for it on fourth down as often as stats suggest they should, compared the problem to a thought experiment called the Ellsberg paradox that shows people lean toward risk aversion when dealing with ambiguous probabilities. In soccer, researchers have suggested that goalkeepers dive for penalties too often due to something called action bias—a fear of looking silly if they stand upright while a ball dribbles into the corner. But punting too conservatively may be smart if a coach knows he could be fired for failing to convert on fourth and short, and keepers’ concern about looking helpless looks less like a bad grasp of math than a good grasp of how fans and coaches think.

There could be a similar thing going on with midfield passing. Although different clubs have different attitudes toward possession, it’s only human for players to get annoyed at a teammate who gives away a low-percentage hero ball. A lot of teams have a sort of “you break it, you bought it” norm where the guy who loses the ball is supposed to bust his ass to win it back, even if it means leaving his position to sprint all the way back to goal. Under social pressure like that, it wouldn’t be surprising if passers turned out to be more risk averse in midfield than they should be. Like with football coaches and diving keepers, trying to rationally maximize your chance of winning isn’t actually the rational thing to do unless you’ve got a culture that supports it.

So should coaches rip up their game models and start playing 60-yard throughballs all the time? I mean yeah obviously if they want to be rad as hell, but this is in no way a declaration of What The Numbers Say. Even if the numbers did say it, any team that built its attack around Neuhaus assists would pretty quickly see its opponents catch on. Instead, think of this as the case for audacity—for trying certain things in part because they’re so unusual—and for thinking twice about getting mad next time they don’t work out. ❧

Further reading:

• David Romer, Do Firms Maximize? Evidence from Professional Football (Journal of Political Economy, 2006)
• Michael Bar-Eli et al., Action Bias Among Elite Soccer Goalkeepers: The Case of Penalty Kicks (Journal of Economic Psychology, 2007)
• Pieter Robberechts, Exploring How VAEP Values Actions

Image: Francisco Goya, Plate 20 from "Tauromaquia": The agility and audacity of Juanito Apiñani in the ring at Madrid